Nguyên lý tạo ảnh của hệ thống chụp X quang kỹ thuật số

Stewart, PhD, DABMP 2 Take Away: Five Things You should be able Take Away: Five Things You should be able to Explain after the DR/Adjuncts Lecture ¬ The various types of detectors used

Trang 1

Brent K Stewart, PhD, DABMP 1

Brent K Stewart, PhD, DABMP Professor, Radiology and Medical Education Director, Diagnostic Physics

a copy of this lecture may be found at:

http://courses.washington.edu/radxphys/PhysicsCourse04-05.html

Take Away: Five Things You should be able

to Explain after the DR/Adjuncts Lecture

¬ The various types of detectors used in digital imaging (e.g., scintillators, photoconductors, etc.)

¬ The differences between the various technologies used for digital radiography (e.g., CR, indirect and direct DR)

¬ Benefits of each type (e.g., resolution, dose efficiency)

¬ Why digital image correction and processing are necessary or useful and how they are executed

¬ The various types of adjuncts to radiology (e.g., DSA or dual-energy imaging), what issue they are trying to resolve, mechanism exploited and end result

Digital Representation of Data (1)

¬ Bits, Bytes and Words

¬ Smallest unit of storage capacity = 1 bit (binary digit: 1 or 0)

¬ Bits grouped into bytes: 8 bits = byte

¬ Word = 16, 32 or 64 bits, depending on the computer system addressing architecture

¬ Computer storage capacity is measured in:

¬ kilobytes (kB) -210bytes = 1024 bytes ≈a thousand bytes

¬ megabytes (MB) -220bytes = 1024 kilobytes ≈a million bytes

¬ gigabytes (GB) -230bytes = 1024 megabytes ≈a billion bytes

¬ terabytes (TB) -240bytes = 1024 gigabytes ≈a trillion bytes

Digital Representation of Data (2)

¬ Digital Representation of Different Types of Data

¬ Alphanumeric text, integers, and non-integer data

¬ Storage of Positive Integers

¬ In general, n bits have 2npossible permutations and can represent integers from 0 to 2n-1 (the range usually denoted with square brackets):

¬ n bits represents 2nvalues with range [0, 2n-1]

¬ 8 bits represents 28= 256 values with range [0, 255]

¬ 16 bits represents 216= 65,536 values with range [0, 65535]

Trang 2

Conversion of Analog Data to Digital Form

¬ The electronic measuring devices of medical scanners (e.g., transducers and detectors) produce analog signals

¬ Analog to digital conversion (analog to digital converter –ADC)

¬ ADCs characterized by

¬ sampling rate or frequency (e.g., samples/sec –1 MHz)

¬ number of bits output per sample (e.g., 12 bits/sample = 12-bit ADC)

c f Bushberg, et al., The Essential Physics of Medical Imaging, 2nd ed., p 69.

Digital Storage of Images

¬ Usually stored as a 2D array (matrix) of data, I(x,y): I(1,1), I(2,1), … I(n,m-1), I(n,m)

¬ Each minute region of the image is called a pixel(picture element) represented by one value (e.g., digital value, gray level or Hounsfield unit)

¬ Typical matrices:

¬ CT: 512x512x12 bits/pixel

¬ CR: 1760x2140x10 bits/pixel

¬ DR: 2048x2560x16 bits/pixel

c.f Huang, HK Elements of Digital Radiology, p 8.

Detectors in Digital Imaging (1)

¬ Gas and solid-state detectors

¬ Energy deposited to e-through Compton and photoelectric interactions

¬ Gas detectors –apply high voltage across a chamber and measuring the flow of e-produced by ionization in the gas (typically high Z gases like Xenon: Z=54, K-edge = 35 keV)

¬ Were used in older CT units

c.f Bushberg, et al The Essential Physics of Medical Imaging, 2 nd ed., p.32.

¬ Solid-state materials

¬ Electrons arranged in bands with conduction band usually empty

¬ Solid-state detectors

¬ Scintillators –some deposited energy converted to visible light

¬ Photoconductors –charge collected and measured directly

¬ Photostimulable phosphors –energy stored in electron traps

c.f Yaffe MJ and Rowlands JA Phys Med Biol 42 (1997), p Elements of Digital Radiology, p 10.

Trang 3

c.f Yaffe MJ and Rowlands JA Phys Med Biol 42 (1997), p Elements of Digital Radiology, p 9.

Periodic Table of the Elements

c.f http://www.ktf-split.hr/periodni/en/

Computed Radiography (CR)

¬ Photostimulable phosphor (PSP)

¬ Barium fluorohalide: 85%

BaFBr:Eu + 15% BaFI:Eu

¬ e-from Eu2+liberated through absorption of x-rays by PSP

¬ Liberated e-fall from the conduction band into ‘trapping sites’ near F-centers

¬ By low energy laser light (700 nm) stimulation the e-are re-promoted into the conduction band where some recombine with the Eu3+

ions and emit a blue-green

(400-500 nm) visible light (VL)

c.f Bushberg, et al The Essential Physics of Medical

Imaging, 2 nd ed., p 295.

Computed Radiography (CR) System (1)

¬ Imaging plate (IP) made of PSP is exposed identically to SF radiography in Bucky

¬ IP in CR cassette taken to CR reader where the IP is separated from cassette

¬ IP is transferred across a stage with stepping motors and scanned

by a laser beam (~700 nm) swept across the IP by a rotating polygonal mirror

¬ Light emitted from the IP is collected by a fiber-optic bundle and funneled into a photomultiplier tube (PMT)

¬ PMT converts VL into e-current

Trang 4

¬ Electronic signal output from PMT input to an ADC

¬ Digital output from ADC stored

¬ Raster swept out by rotating polygonal mirror and stage stepping motors produces I(t) into PMT which eventually translates into the stored DV(x,y):

PMT ADC RAM

¬ IP exposed to bright light to erase any remaining trapped e-(~50%)

¬ IP mechanically reinserted into cassette ready for use

¬ 200µm and 100µm pixel size -(14”x17”: 1780x2160 and 3560x4320, respectively)

¬ IP dynamic range = 104, about 100x that of S-F (102)

¬ Very wide latitude flat contrast

¬ Image processing required:

¬ Enhance contrast

¬ Spatial-frequency filtering

¬ CR’s wide latitude and image processing capabilities produce reasonable OD or DV for either under or overexposed exams

¬ Helps in portable radiography:

where the tight exposure limits of S-F are hard to achieve

¬ Underexposed quantum mottle and overexposed unnecessary patient dose

Charged-Coupled Devices (CCD)

¬ Form images from visible light

¬ Videocams & digital cameras

¬ Each picture element (pixel) a photosensitive ‘bucket’

¬ After exposure, the elements electronically readout via ‘shift-and-read’ logic and digitized

¬ Light focused using lenses or fiber-optics

¬ Fluoroscopy (II)

¬ Digital cineradiography (II)

¬ Digital biopsy system (phosphor screen)

¬ 1K and 2K CCDs used

c.f Bushberg, et al The Essential Physics of Medical Imaging, 2 nd ed., pp 298-299.

Indirect Flat Panel Detectors

¬ Use an intensifying screen (CsI) to generate VL photons from an x-ray exposure

¬ Light photons absorbed by individual array photodetectors

¬ Each element of the array (pixel) consists of transistor (readout) electronics and a photodetector area

¬ The manufacture of these arrays is similar to that used in laptop screens: thin-film transistors (TFT)

Trang 5

Thin-Film Transistors (TFT)

¬ After the exposure is complete and the e-have been stored in the photodetection area (capacitor), rows in the TFT are scanned, activating the transistor gates

¬ Transistor source (connected

to photodetector capacitors is shunted through the drain to associated charge amplifiers

¬ Amplified signal from each pixel then digitized and stored

¬ X-ray VL e- ADC RAM

Resolution and Fill Factor

¬ Dimension of detector element largely determines spatial resolution

¬ 200µm and 100µm pixel size typical

¬ For dimension of ‘a’ mm -Nyquist frequency: FN = 1/2a

¬ If a = 100µm FN = 5 cycle/mm

¬ Fill factor = (light sensitive area)/(detector element area)

¬ Trade-off between spatial resolution and contrast resolution

c.f Bushberg, et al The Essential Physics of Medical Imaging, 2 nd ed., p 303.

Direct Flat Panel Detectors

¬ Use a layer of photoconductive material (e.g., -Se) atop a TFT array

¬ e-released in the detector layer from x-ray interactions used to form the image directly

¬ X-ray e- TFT ADC RAM

¬ High degree of e-directionality through application of E field

¬ Photoconductive material can be made thick w/o degradation of spatial resolution

¬ Photoconductive materials

¬ Selenium (Z=34)

¬ CdTe, HgI 2 and PbI 2

Indirect Flat Panel Detector (for comparison)

Direct Flat Panel Detector

Periodic Table of the Elements

c.f http://www.ktf-split.hr/periodni/en/

Trang 6

Digital versus Analog Processes & Implementation

¬ Although some of the previous image reception systems were labeled ‘digital’, the initial stage of those devices produce an analog signal that is later digitized

¬ CR: x-rays VL PMT current voltage ADC

¬ CCD, direct & indirect digital detectors: stored e- ADC

¬ Benefits of CR

¬ Same exam process and equipment as screen-film radiography

¬ Many exam rooms serviced by one reader

¬ Lower initial cost

¬ Benefits of DR

¬ Throughput : radiographs available immediately for QC & read

Patient Dose Considerations

¬ Over and underexposed digital receptors produce images with reasonable OD or gray scale values

¬ As overexposure can occur, need monitoring program

¬ CR IP acts like a 200 speed S-F system wrt QDE

¬ Use the CR sensitivity (‘S’) number to track dose

¬ Bone, spine and extremities: 200

¬ Chest: 300

¬ General imaging including abdomen and pelvis: 300/400

¬ Flat panel detectors can reduce radiation dose by 2-3x

as compared with CR for the same image quality due to quantum absorption efficiency & conversion efficiency

S-number Dashboard Main Exams

0 50 100 150 200 250 300 350 400 450 500

Exam Code

Sep-03 Baseline Mar-04 Apr-04

Target Values

Fixed Chest - [255-345]

Bone, Spine & Ext -[170-230]

General Imaging incl

Using the CR Sensitivity Number to Track Dose

Huda Ch6: Digital X-ray Imaging Question

¬ 12 Photostimulable phosphors do NOT include:

¬ A Analog-to-digital converters

¬ B Barium fluorohalide

¬ C Light detectors (blue)

¬ D Red light lasers

¬ E Video cameras

Trang 7

¬ 11 Which of the following x-ray detector materials emits visible light:

¬ A Xenon

¬ B Mercuric iodide

¬ C Lead iodide

¬ D Selenium

¬ E Cesium iodide

Raphex 2002 Question: Digital Radiography

¬ D47 Concerning computed radiography (CR), which of

the following is true?

¬ A Numerous, small solid-state detectors are used to capture the x-ray exposure patterns

¬ B It has better spatial resolution than film

¬ C It is ideal for portable x-ray examinations, when phototiming cannot be used

¬ D It is associated with high reject/repeat rates

¬ E The image capture, storage, and display are performed by the receiver.

¬ 13 Photoconductors convert x-ray energy directly into:

¬ A Light

¬ B Current

¬ C Heat

¬ D Charge

¬ E RF energy

Digital Image Correction

¬ Interpolation to fill in dead pixel and row/column defects

¬ Subtracting out average dark noise image Davg(t)(x,y)

¬ Differences in detector element digital values for flat field

¬ Gain image: G(x,y) =G’(x,y) -Davg(t)(x,y); Gavg=(1/N)·∑∑G(x,y)

¬ Make corrections for each detector element (map)

¬ I(x,y) = Gavg· Iraw(x,y) -Davg(t)(x,y)] / G(x,y)

¬ Done for DR and in a similar manner for CT (later)

¬ Not performed for CR on a pixel by pixel basis, although there are corrections on a column basis for differences in light conduction efficiency in the light guide to the PMT

Trang 8

Digital Image Correction

Global Processing

¬ Most common global image processing: window/level

¬ Global processing algorithm

¬ I’(x,y) = c · [I(x,y) –a]:

essentially y = mx+ b

¬ Level (brightness) set by a

¬ Window (contrast) set by c

¬ I’ = [2N/ww]·[I-wl-(ww/2)}], where ww= window width and

wl= window level

¬ Need threshold limits when max/min [2N-1, 0] digital values encountered

¬ If I’(x,y) > Tmax I’(x,y) = Tmax

¬ If I’(x,y) < Tmin I’(x,y) = Tmin

c.f Bushberg, et al The Essential Physics of Medical Imaging, 2 nd ed., pp 92 and 311.

Image Processing Based on Convolution

¬ Convolution: Ch 10 -Image Quality and Ch 13 -CT

¬ Defined mathematically as passing a N-dimensional convolution kernel over an N-dimensional numeric array (e.g., 2D image or CT transmission profile)

¬ At each location (x, y, z, t, ) in the number array multiply the convolution kernel values by the associated values in the numeric array and sum

¬ Place the sum into a new numeric array at the same location

¬ Delta function kernel

¬ Blurring kernel (normalization) also known as low-pass filter

¬ Edge sharpening kernel

0 0 0 0 1 0 0 0 0

1/9 1/9 1/9 1/9 1/9 1/9 1/9 1/9 1/9

-1 -1 -1 -1 9 -1 -1 -1 -1

Trang 9

¬ Convolution kernels can be much larger than 3 x 3, but usually N x M with N and M odd

¬ Can also perform edge sharpening by subtracting blurred image from original high-frequency detail (harmonization)

¬ The edge sharpened image can then be added back to the original image to make up for some blurring in the original image: CR unsharpmasking - freq processing

¬ The effects of convolution cannot in general be undone

by a ‘de-convolution’ process due to the presence of noise, but a deconvolution kernel can be applied to produce an approximation: 19F MRI

Median and Sigma Filtering

¬ Convolution of an image with a kernel where all the values are the same, e.g (1/NxM), essentially performs

an average over the kernel footprint

¬ Smoothing or noise reduction

¬ This can make the resulting output value susceptible to outliers (high or low)

¬ Median filter: rank order values in kernel footprint and take the median (middle) value

¬ Sigma filter: set sigma (σ) value (e.g., 1) and throw out all values in kernel footprint > µ + σ or < µ – σ and then take the average and place in output image

Multiresolution/Multiscale Processing and Adaptive Histogram Equalization (AHE)

¬ Some CR systems (Agfa/Fuji) make use of multiresolution image processing (AKA unsharpmasking)

to enhance spatial resolution

¬ Wavelet or pyramidal processing on multiple frequency scales

¬ Histogram equalization re-distributes image digital values to uniformly span the entire digital value range [2N-1,0] to maximize contrast

¬ AHE does this on a spatial sub-region basis in an image rather than the entire image

¬ Fuji ‘Dynamic Range Control’ (DRC) a version of AHE that operates on sub-regions of digital values

Trang 10

Histogram Equalization

c.f http://www.wavemetrics.com/products/igorpro/imageprocessing/imagetransforms/histmodification.htm

Global and Adaptive Histogram Equalization

The following images illustrate the differences between global and adaptive histogram equalization.

MR image with the corresponding gray-scale histogram The histogram has a peak at minimum intensity consistent with the relatively dark nature of the image

Global histogram equalization and the final gray-scale histogram Comparing the results with the figure above we can see that the distribution was shifted towards higher values while the peak at minimum intensity remains.

Adaptive histogram equalization shows better contrast over different parts of the image The corresponding gray-scale histogram lacks the mid-levels present in the global histogram equalization as a result of setting a high contrast level.

c.f http://www.wavemetrics.com/products/igorpro/imageprocessing/imagetransforms/histmodification.htm

Contrast vs Spatial Resolution in Digital Imaging

¬ S-F mammography can produce images w/ > 20 lp/mm

¬ According to Nyquist criterion would require 25 µm/pixel resulting in a 7,200 x 9,600 image (132 Mbytes/image)

¬ Digital systems have inferior spatial resolution

¬ However, due to wide dynamic range of digital detectors and image processing capabilities, digital systems have superior contrast resolution

Digital Imaging Systems and DQE

¬ Remember the equation for DQE(f)?

¬ DQE(f) =

¬ How can we account for this?

¬ Both CR and the screens in film/screens made thin

¬ Film higher spatial resolution than CR

¬ DQE higher for -Si systems using CsI and Gd2O2S rather than -Se (mean x-ray E & Z)

¬ -SiDQE falling off more rapidly than -Se (geometry)

( ) ( )

k MTF f

N NPS f ⋅

-Si DR -Se DR

Định dạng
Số trang	13
Dung lượng	0,95 MB